I have a driver that raises some warnings/errors during compilation, since the proc_fs api changed since its creation. The driver still uses create_proc_entry while the latest api version I am aware off, offers proc_create. Since I am new to driver programming under linux I tried to look at the source, however my ctags skills must be lacking since I only found proc_create in proc_fs.h. However, I would like to look at the implementation or some documentation, to know what error codes it returns so I know what I have to handle.
Can you point me to documentation for the proc api, or the source holding the definition of proc_create? A hint how it can be found would be appreciated as well.
create_proc is defined as an inline function, so it is fully implemented in the header proc_fs.h. It basically calls proc_create_data with a NULL for the data argument.
There seems very little documentation on these functions in the source so I'd recommend looking at other call sites within the kernel source, which you cam see listed in this LXR search.
BTW, creating new files in /proc seems frowned upon these days - it seems sysfs is where new interfaces should be created.
Best documentation for an API would be existing code itself. Check a few c files in fs/proc or grep for proc_create in the source code
Related
Assume you get a new ebpf program type BPF_PROG_TYPE_SOCKET_FILTER， i want to get knowledge for this program type, for example:
what do I do with this program type?
how do I attach my BPF program for this program type?
when does the attached program get run?
what context is provided to my program?
There is no documents describing this information, should i read kernel source code? where should i start?
There is a great blog https://blogs.oracle.com/linux/post/bpf-a-tour-of-program-types , but i'm interesting at how to get this information by myself, because there are always new program type exists.
Unfortunately, there's no simple answer. There's an effort beginning in the BPF community to bring documentation and standardisation for eBPF, but it's not there yet. In the meantime, if you want to dig the info by yourself, this means one thing: reading kernel code.
what do I do with this program type?
This is usually introduced in the commit description for the commit (or the patchset) that added the feature. You can usually find the relevant commit by git blame-ing the new type in the UAPI header, include/uapi/linux/bpf.h. It's also worth looking at the samples (samples/bpf/) or tests (tools/testing/selftests/bpf/) using the programs of a given type to see how and what they do with it.
how do I attach my BPF program for this program type?
In many cases, libbpf or other libraries will handle this for you. Otherwise, looking at the tests and samples would tell.
when does the attached program get run?
You need to look at the code. This is generally added in the commit that brought the new type as well. The name of the macro or function used to call the BPF programs has changed over time.
what context is provided to my program?
Again, you can usually tell from the description of the feature and/or the examples.
I'm trying to remove a data source from a rrd db.
I found I can do something like
"rrdtool tune mydb.rrd DEL:source_name"
and it works, but I want to do it from C/C++ code.
I could use the system function in Linux, but I don't
like the overhead.
I looked in https://oss.oetiker.ch/rrdtool/doc/librrd.en.html
to see if there is something I could use, but I didn't find anything.
I also looked in the rrd source code from https://github.com/oetiker/rrdtool-1.x/tree/master/src
and I found they call rrd_modify_r2() to remove sources, but this function is static, so it's not exported (as opposed to rrdc_create_r2)
So, how can I remove a source from C/C++ code ?
thanks,
Catalin
You use of course rrdtool tune filename.rrd DEL:ds-name to do this from the commandline, as you note.
However the RRDTool C bindings in librrd are not as comprehensive, and do not appear to expose this functionality. Not sure why - the modify functions is clearly useful - but thats how it seems to be.
One option you have would be to simply call the external rrdtool binary with fork/exec, passing the appropriate commandline. This is not a particularly pretty way to do it, but is more portable and compatible with the published interface.
Can anyone please let me know any pointers to create Sparse files(holey files) in latest vdbench 50404rc2. It seems that this is the latest supported feature.
Link for more info:
https://community.oracle.com/thread/3759500?start=0&tstart=0
The answer was given by Henk, but on oracle vdbench forum so posting the excerpt from it,also the below link to forum post.
This is EXPERIMENTAL, so it will work for now, but once I get feedback and decide that this experiment was successful I will change the instructions to activate it.
That means that the '-d86' info below will no longer work.
To activate truncate, add '-d86' as an execution parameter, or, add 'debug=86' at the top of your parameter file.
(For experiments adding an other 'debug=' parameter is much easier than fiddling with the Vdbench parameter parser. If I decide to make this permanently available I'll worry about adding a more 'official' parameter.)
This uses the Unix 'ftruncate()' and similar Windows function during file creation.
This will create ONLY sparse files during the format, not one block of data is written until further Vdbench workloads are run against these files.
The definition file.
debug=86
fsd=fsd1,anchor=/tmp/sparsedir,depth=1,width=1,files=10,size=40k
fwd=fwd1,fsd=fsd1,operation=read,xfersize=4k,fileio=sequential,fileselect=random,threads=1
rd=rd1,fwd=fwd1,fwdrate=max,format=yes,elapsed=10,interval=1
I am trying to implement a spin lock using the test_and_set_bit function. I found a bitops.h file which consisted of this function. However, in my current kernel version which is 3.0, the function is not included in that header file i.e, bitops.h. Any anyone provide some references where I can find that?
Not sure if I totally understand your question, but including <linux/bitops.h> should bring in the definition of test_and_set_bit(). The actual definition of the function is not in include/linux/bitops.h but it is picked up via the include of <asm/bitops.h> that is in the linux/ version of the include.
So to see the actual definition of test_and_set_bit() you can look in arch/arm/include/asm/bitops.h or arch/x86/include/asm/bitops.h (or whatever other architecture you're interested in).
By the way, there's no reason to need to implement your own spinlock -- the kernel has (of course) the standard spinlock_t and also functions like bit_spin_lock() that use a single bit as a lock.
I need a set of key-value pairs for configuration read in from a file. I tried using show on a Data.Map and it doesn't look at all like what I want. It seems this is something many others might have already done so I'm wondering if there is a standard way to do it and what library to use.
Go to hackage.
Click on "packages"
Search for "config".
Notice ConfigFile(TH), EEConfig, and tconfig.
Read the Haddock documentation
Select a couple and implement your task.
Blog about your findings so the rest of us can learn from your new found expertise (thanks!).
EDIT:
I've recently used configurator - which was easy enough. I suggest you try that one!
(Yes, yes. If I took my own advice I would have made a blog for you all)
The configuration category on Hackage should list all relevant libraries:
http://hackage.haskell.org/packages/#cat:Configuration
I have researched the topic myself now, and my conclusion is:
configurator is very good, but it's currently only for user-edited configurations. The application only reads the configuration and cannot modify it. So it's more for server-side applications.
tconfig has a a simple API and looked like it was what I wanted, maybe a bit raw, until I realized it's unmaintained and that some commits which are really important to use the app are applied on github but the hackage package was not updated
Other solutions didn't look like they'd work for me, I didn't like the API, but every application (and tastes) are different.
I think using JSON for instance is not a good solution because at least with Aeson when you add new settings in a new release, the old JSON without the new member from the previous version won't load. Also, i find that solution a bit verbose.
The conclusion of my research is that I wrote my own library, app-settings, which aims to be key-value, read-write, with a as succint and type-safe API as possible. And you'll find it also in the hackage links for the configurations category that I gave.
So to summarize, I think configurator is the standard for read-only configurations (and it's very powerful too, you can split the configuration file with imports for instance). For read-write there are many small libraries, some unmaintained, and no real standard I think.
UPDATE 2018 be sure to look at dhall
I'd also suggest just using Text.JSON or one of the yaml libraries available (I prefer JSON myself, but...).
The configfile package looks like what you want.